Welcome to the 28th issue of Last Week in AWS.
re:Invent draws steadily closer. More about that to come in the next few weeks; more germane to the next couple of weeks, I’ll be speaking at DevOops next week, and All Things Open the week after– if you’re near either of these, please let me know; I’d love to catch up with readers.
Community Contributions
One of the best parts of re:Invent is the ability to meet with a variety of movers and shakers within the ecosystem surrounding Amazon, Gartner’s Public Cloud Infrastructure as a Service leader with respect to both vision and execution ability– okay, for those of you still paying attention and didn’t skip ahead to less dry stuff, here’s your guide to re:Invent Parties. I’m looking to seeing some of you there!
Persisting state between spot instances was partially solved by the ability to stop and resume introduced last month– but it’s not a panacea. This blog post goes into an alternate approach involving shoving everything into containers.
This happened a couple of weeks ago, but I missed it until now– so this week’s S3 Bucket Negligence award goes to SVR Tracking. While you’ve probably never heard of them, they just leaked logins for over half a million car tracking devices. Bravo. There’s no way that’s going to end poorly.
This reads like a sales pitch and I don’t even care– they’ve built a service to let you write mainframe code in COBOL on AWS. Bonus points for using the phrase “Mainframe DevOps” without even a trace of sarcasm. Please shine on, you crazy diamonds.
While I’ll admit that AWS Does Not Protect You From DevOps is a title that has a lot of comic possibilities, the author’s point is well founded: nothing in AWS is simplistic, the edge cases will cut you to death if you’re unwary, and the “Amazon is going to take systems people’s jobs away!” hysteria is no closer to being true now than it was in 2008. If “click button, receive DevOps” were a thing, I wouldn’t have a job– much less this newsletter.
An AWS developer demonstrates using Lambda for rapid computation of 305 million solutions to the Black-Scholes model. The model itself is used to calculate the price of European equities, so it’s of great interest to financial mathematicians, as well as SoundGarden– who can’t resist a Black-Scholes Hun.
Another week, another glorious drawing of an AWS service from awsgeek; this week’s features AWS Lambda.
An Amazon spokesperson went on record this week to rebut Oracle founder Larry Ellison’s incorrect assertions about Redshift’s elastic capabilities. I would like to grab a drink and talk snark with that spokesperson if they’re at re:Invent this year— if you know who they are, please reach out to them and let them know where to find me. I promise anonymity.
Amazon CTO Werner Vogels reflects upon ten years of Dynamo, and its eventual launch as the product DynamoDB. If you read it carefully, you’ll notice that there is absolutely no mention either here nor in the original post ten years ago of SimpleDB. Somewhere, a single tear falls down the SimpleDB product owner’s cheek.
The NextDoor engineering blog talks about Bender, their new Java framework for building ETL jobs on top of Lambda. Needs more Futurama references…
Lynn Langit is legendary in the cloud space– not only is she an AWS “Community Hero,” she’s also a Google Cloud Developer Expert and a former Microsoft employee. In other words, she’s professionally versatile and/or confused. In this edition of the “Serverless Heroes” series by A Cloud Guru, she opines on the future of containers (“I don’t care”), GCP’s cultural dysfunction as a platform, and acknowledges that Serverless has one heck of a learning curve.
Choice Cuts From the AWS Blog
AWS VPN Update – Custom PSK, Inside Tunnel IP, and SDK update – VPNs now include a variety of– okay, it’s not cross-region VPC peering. Until it is, nobody is going to care about anything in this space, Amazon. I’m sorry to be the one to break the news to you, but it’s 2017. Get with the program please.
Amazon EMR now supports per-second billing – It’s time to revisit the economics of using spot fleets for EMR workloads– and potentially your life choices if that sentence makes sense to you.
Amazon DynamoDB Accelerator (DAX) releases SDK for Node.js, support for R4 instances and general availability in Asia Pacific (Mumbai) and South America (São Paulo) regions – A bunch of changes dropped this week for DynamoDB and its accelerator, including regions instance sizes, a javascript SDK, and a continuing lack of a tacit admission that most of the feature enhancements to DynamoDB over the past year have been centered around making a shaky offering more acceptable to work with. Good work– but this is all evolutionary rather than revolutionary. re:Invent is around the corner– impress us!
AWS Database Migration Service Adds Amazon S3 and Azure SQL Database as Sources – DMS now supports additional sources. No word yet on whether it munges your schema for those sources like it did many at launch. I get that early release is baked into Amazon’s DNA, but databases and filesystems are two areas in which it pays to be technically conservative.
Improved Testing on the AWS Lambda Console | AWS Compute Blog – Amazon is pleased to announce that there are better ways of testing your Lambda functions, as well as beg the question of whether you test your Lambda functions at all.
Tools
If you’re using MongoDB, you may be interested in how to back up your MongoDB databases to S3.
For those Mac users who spend a lot of time in the AWS console, you might take a look at this tool that launches AWS console sections from your dock.
Use MacOS and a variety of AWS accounts with MFA enabled? Need to assume roles between those accounts? For the three people left with their hands raised, check out aws-assume-role.
Maybe if I recommend enough tools to check S3 bucket ACLs, I’ll stop seeing leaky S3 buckets. Here’s this week’s.
A Lambda function that rips text from a variety of binary formats seems like a handy thing to me. Please do something hilarious with this, then write up something I can link against.
Tip of the Week
This was more relevant before you were able to expand VPCs, but consider non-NLB scaling when you’re sizing your subnets. ELBs will launch ENIs in your subnets and use up various IPs. This probably won’t matter to you at all until you’re suddenly no longer able to launch new things in those subnets, and your infrastructure catches fire.
Note that NLBs don’t have IP endpoints in your subnets– they’re weird like that.
…and that’s what happened Last Week in AWS.
